Geneticalgorithms And Itsapplication On Constrained Optimization

Nowadays complex constraint optimization problems (COPs) keep emerging in the field of engineering, management, economic, scientific research and military. These problems which are solved by intelligent optimization methods generally are always large-scale, highly complex, uncertain and difficulty in modeling. Genetic algorithm (GA), which imitates natural organisms’ evolutionary process and mechanism, is a kind of intelligent optimization algorithm with the advantages - operability, high degree of parallel, adaptability and wide application. Genetic algorithm has become a hot issue in the field of artificial intelligence. This thesis mainly studies coding methods of genetic algorithm, genetic operators, and GA’s application on complex constraint optimization problems. Main content can be summed up in the following four aspects:1. To solve a type of nonlinear programming and nonlinear integer programming problems, a genetic algorithm based on a new real code (NRCGA) is proposed. The new real coded strategy is adopted and all of the infeasible chromosomes generated after genetic operation can be repaired by simply sorting. The novel algorithm, which can handle a kind of constraints independently and solve complex constrained optimization problems combining with other constraint-handling techniques, is a method based on decoders with no additional parameters. Six examples show that the new algorithm possesses high search efficiency and strong robustness. 2. To solve convex quadratic knapsack problem (QKP), an improvedvariable-grouping based genetic algorithm is proposed. The VGGA first eliminates part of decision variables based on the optimal solution to the QKP’s continuous relaxation problem, and then applies VGGA to the subproblem after part of variables are eliminated. So, QKP’s optimal solution can be found. Numerical examples show that the improved algorithm is superior to VGGA.3. A hash function designinng method based on GA is proposed. According to analyse the similarity of folding method and division hashing method, the hash function designinng problem is converted into an integer programming problem, the optimal solution of which can be solved by NRCGA. Finally, the hash function is constructed according to this solution. Numerical examples show that the new method is effective.4. A novel constrained optimization evolutionary algorithm based on decoders is proposed. Applying GA as evolution algorithm, the novel approach builds an one-one correspondence between an arbitrary n-dimension vector and its unit vector and modulus so as to construct a real-code whose redundancy is zero. Experiments on 13 benchmark test problems and three frequently-used test problems verify the effectiveness and ef?ciency of the proposed method.